A deep Generative Artificial Intelligence system to predict species coexistence patterns

Abstract

Predicting coexistence patterns is a current challenge to understand diversity maintenance, especially in rich communities where these patterns' complexity is magnified through indirect interactions that prevent their approximation with classical experimental approaches. We explore cutting‐edge Machine Learning techniques called Generative Artificial Intelligence (GenAI) to predict species coexistence patterns in vegetation patches, training generative adversarial networks (GAN) and variational AutoEncoders (VAE) that are then used to unravel some of the mechanisms behind community assemblage. The GAN accurately reproduces real patches' species composition and plant species' affinity to different soil types, and the VAE also reaches a high level of accuracy, above 99%. Using the artificially generated patches, we found that high‐order interactions tend to suppress the positive effects of low‐order interactions. Finally, by reconstructing successional trajectories, we could identify the pioneer species with larger potential to generate a high diversity of distinct patches in terms of species composition. Understanding the complexity of species coexistence patterns in diverse ecological communities requires new approaches beyond heuristic rules. Generative Artificial Intelligence can be a powerful tool to this end as it allows to overcome the inherent dimensionality of this challenge.

Plant metabolic response to stress in an arid ecosystem is mediated by the presence of neighbors

“…Similarly, closely related plant species tend to share pests and pathogens (Gougherty & Davies, 2021) but may also have similar defensive chemistry (Agrawal, 2007). Finally, the absence of a relationship between the metabolic strategy and phylogenetic diversity of the neighborhood may come from the limited ability of paired phylogenetic distances to capture the complexity of indirect interactions occurring within the study vegetation patches (Hirn et al, 2022).…”

Defining model complexity: An ecological perspective

“…For example, a statistical model such as a linear regression has at least two parameters: an intercept and a slope. Model complexity increases with each additional parameter (either including more covariates or representing interactions between covariates), up to hundreds (e.g., general circulation models predicting global climate patterns; Dunne et al, 2012) or even millions of parameters (e.g., >10 million parameters in large language models and other deep learning models; Hirn et al, 2022; Rostami et al, 2023). In one application of parameter complexity, Clark et al (2020) used the number of parameters to demonstrate that a model of intermediate complexity (i.e., an intermediate number of parameters) produced the best out‐of‐sample predictions of species abundances within a grassland plant community.…”

Deep learning as a tool for ecology and evolution

“…This involves using generative adversarial networks (GANs; Box 1) (Wang, She, & Ward, 2019) to create artificial human genomic sequences of known ancestry (Montserrat et al, 2019; Yelmen et al, 2021). GANs have also been leveraged to simulate realistic population genetic data for inference of population genetic parameters (Wang, Wang, et al, 2021), to augment training data with artificial images (Klasen et al, 2021; Lopes et al, 2021) and model vegetation succession to gain insight into species interactions (Hirn et al, 2022).…”